RFD 1150: Shade colour is solved per tone, not multiplied
check_anime_materials.py, anime-materials.cff
Problem
MToon extends glTF pbrMetallicRoughness rather than replacing it, so the lit colour is baseColorFactor and the extension adds a shade colour nothing in PBR constrains. Somebody has to choose it.
The obvious choice, one multiplier for every tone, fails on equity. At x0.6 the lit-to-shade separation is dE 17.28 at Monk 1 and dE 4.83 at Monk 10, a factor of 3.6. The terminator is the shading cue a detector reads, so one multiplier builds a detection gap into the corpus that would later present as a model defect.
Decision
The multiplier is solved per tone to hold the separation at dE 12. anime_materials.usda stores the tone identities and the target and nothing else, because the multiplier, the shade colour and the achieved separation all follow from those.
The dark end is bounded and reported rather than clamped: dE 12 at Monk 10 needs shade L* 2.9, and the gate prints what each tone reached.
Applying a scale defined on photographs of real skin to a toon material is a substitution, recorded as one. No anime skin-tone standard exists.
References
DETAILS.mdcarries the ladder and the solved multipliers.